X2 4400+ OC to 4800+ Requires Slow Hypertransport Bus?

Hurin

2[H]4U
Joined
Oct 8, 2003
Messages
2,410
Hi All,

I'm a former Intel weenie about to take the plunge into AMD-ville.

At first, I was considering the X2 4800+. But it seems like a relatively easy overclock to get the 4400+ up to that speed and save the $300. But I'm concerned because it looks like I'll need to turn the hypertransport bus down below 1000MHz in order to overclock the CPU by the desired 200MHz.

Here are the parts I'm considering:

X2 4800+ Toledo
Asus A8N-SLI Premium
Corsair TWINX2048-3200C2

Now, the CPU multiplier is locked at 11. So, to get to 2.4GHz (4800+ speeds), I'll need to raise the FSB to 219MHz. But this will make my Hypertransport bus run at 1095MHz. The only cure for that, of course, is to change the multiplier for the Hypertransport bus to 4. But then I'm running the hypertransport bus at 876MHz.

Will running the Hypertransport bus at 876MHz noticeably harm overall system performance?

Of course, it would be great if the CPU could handle a 250fsb (CPU at 2.75GHz!) which would bring the hypertransport bus back to 1000MHz with the multiplier of 4 . But I doubt I can pull that off on only air cooling for the CPU. And air is the only option for me right now.

It looks like, where the memory bus is concerned, I'm stuck at 1:1 ratio. But I think the Corsair C2 sticks are up to only 219MHz at decent timings.

So, I guess my only real concern is what I put in bold. But if some of you AMD gurus want to take a look at all this and tell me if I'm making some terrible mistakes, that would also be much appreciated? Basically, do I understand all this right?

Thanks!

Hurin
 
A HTT link at 876 is plenty of bandwidth. Indeed, the first generation of Opterons only supported 800MHz HTT, and Opterons are designed for servers and workstations that require quite a bit more expansion bus bandwidth than a desktop ever will. Even at 876MHz, you're still cooking along at about 7GB/sec for HT bandwidth. PCI-Express X16 is 4.0GB/sec, 32bit PCI is .133 GB/sec/ There will be plenty of bandwidth to go around even after you OC.
 
Don't know the answers to most of your questions, but I figured I'd chime in because I've read in many forums that the 4400+ X2 isn't that great of an overclocker. Many people have found that one core would O/C well, while the other didn't and this held them back.

Best bang for the buck right now seems to be the 3800+ X2. People have been getting it to 2.75 on air.

.........just some food for thought. :D
 
Wellllll..... it's possibly you could get away with running it at 1095... I have seen folks run it up to 1200, believe it or not... If that does not work, just you might (maybe?) be able to run your cpu faster that you extpected/anticipated in which case you can lower the htt to 4... I am running mine at 245 x 4 so 980 effective... fsb @ 245 and cpu multiplier @ 10. 2450mhz. I can get my f'ing winchester to go faster on this darn MSI board. foo!! anyways... but I should not really comment on the OC'ability of the X2 cores as I have not read up on those too much. I think 876 should be fine though... I have seen folks run theirs @ 850 and stuff without complaint.
 
I run mine at 283x3 (849). I don't see any difference even if I run it at 283x2 (566)
 
Thanks for the replies guys. I'll of course give the near-1100MHz ht-bus a try before clocking it down.

Scroatdog, I've read some rumblings about the X2 4400+ not being a good overclocker as well. But, I'm only asking 200MHz from it. . . anything else would be gravy. I'd say I have an 80% or higher chance of eeking out 200MHz on air, based on what I've read. Sound reasonable?

But your point about the X2 3800+ is well taken and I'm tempted to try the "spend less, get more" approach there if I would conceivably end up with a higher clock speed on air for another $200 less. That $200 savings plus the $300 I'm saving by not getting the X2 4800+ can pay for my second 7800GTX! I think, however, I have a bit of tunnel-vision where these Toledo cores are concerned. For some reason, I really want that 1MB L2 cache!

Again, thanks for the replies guys. It's comforting just to know that I didn't make any bone-headed AMD-newbie mistakes in my math or terminology. At least, none that you've pointed out! :)

H
 
Theres been benchmarks done on HTT speed before, and even a HTT speed of 400MHz doesnt really effect performance, so honestly, you have nothing to worry about.
 
Athlon 64's are not starved for bandwidth with HTT. I guess the best way to give you an idea of how it works, is to think of it like this:

A 3 foot wide pipe with a garden hoses worth of water flowing through it is not gonna be restricted by lowering the width of the pipe to 2 feet.

Since your athlon 64 isnt using all the memory bandwidth it has available, lowering the HTT will not affect performance since the cpu isnt utilizing the maximum amount of bandwidth anyway.
 
Okay, so the hypertransport bus speed question is well and truly answered. Thanks guys. Looks like, even with SLI 7800GTXs, it won't even come near to saturating it.

Now, if I could only decide between the X2 4400+ and 3800+. Darn you Scroatdog! you have me bargain shopping again!
 
2uantuM said:
Athlon 64's are not starved for bandwidth with HTT. I guess the best way to give you an idea of how it works, is to think of it like this:

A 3 foot wide pipe with a garden hoses worth of water flowing through it is not gonna be restricted by lowering the width of the pipe to 2 feet.

Since your athlon 64 isnt using all the memory bandwidth it has available, lowering the HTT will not affect performance since the cpu isnt utilizing the maximum amount of bandwidth anyway.

The HTT bus and memory bus are different on the A64 AFAIK. This is why it doesnt need much HTT speed, because the memory bandwidth doesnt go via the standard HTT bus.
 
Well, look at it this way. You'd save yourself some money to tuck away on maybe a watercooling setup if you don't have one already? Or tuck it away for future upgrades because, as you and everyone else here know, THIS WILL NEVER STOP ! There will always be something newer and better out there and we all want it. Question is, do we attack it from a semi-cost conscious standpoint or not??

Anyhow, I think by getting a 3800+ X2 up to 2.7 from 2.0, the benefits of the extra L2 cache don't really matter anymore.

I'll direct you to these threads about the O/C'ing of the 3800+. Enjoy !

OC Forums

Xtreme Forums 1

3800 x2 on Phase

3800 X2 Under Dry Ice

YMMV widely on the last two threads, though. :p
 
Scroatdog said:
I'll direct you to these threads about the O/C'ing of the 3800+. Enjoy !

OC Forums

Xtreme Forums 1

3800 x2 on Phase

3800 X2 Under Dry Ice

YMMV widely on the last two threads, though. :p

Heh! Already found you over there! :)

Question though, I keep reading where some of those guys are constantly going back to "burn in" their CPUs at lower clock speeds overnight before trying another round of Overclocking the following day. Is there any science to this at all? I mean, beyond the initial few hours of CPU usage, is there really any value to "burning in a CPU" in order to get a higher overclock?

Thanks,

H

Edit: To be more clear, it seems like I'm seeing a lot of people claiming on those links that going back and "burning in" their CPU now allows them to get a higher OC than they were able to do before.
 
I've run mine at 1300 before. Didn't have a problem. No increase in performance either. Instead I returned it to stock and undervolted it since the stock speed handles everything I need right now and I can always overclock later (I hit 2.6ghz on air). But mine's a single core. I'd love a dual core...but it's not in my budget now or anytime in the near future :(
 
Hurin said:
Question though, I keep reading where some of those guys are constantly going back to "burn in" their CPUs at lower clock speeds overnight before trying another round of Overclocking the following day. Is there any science to this at all? I mean, beyond the initial few hours of CPU usage, is there really any value to "burning in a CPU" in order to get a higher overclock?

Thanks,

H

Edit: To be more clear, it seems like I'm seeing a lot of people claiming on those links that going back and "burning in" their CPU now allows them to get a higher OC than they were able to do before.

Yeah, I noticed the same thing. I guess there is something to it, then. I didn't do that when I O/C'd my P4 3.0 I tried 230, 240, 245, and then 250 FSB and then left it there.
 
i have my x2 4400+ doing 250x4 10muitplier giving me 2500mhz
the ram is at 1.5-2-2-5

i never noticed a diff when i had the htt x2 or 3 for that matter


GL!
 
Back
Top